ENFR
8news

Tech • IA • Crypto

TodayShortsTop StoriesTopicsAll videosYT channelsCryptoArchivesFavorites

France bans under-15 social media; Kimi K3 vs GPT-5.6 race

AISaturday, August 1, 2026· 8 videos

Briefing

Audio player
0:00 / 0:00

France bans under-15 social media

France passed a law on 21 July prohibiting social media access for users under 15, the first such national rule in the EU. The measure won broad cross-party support and targets major platforms operating in the country. Enforcement now hinges on robust age checks that platforms must implement. The law positions France as a test case for stricter youth protections online across Europe.

Identity checks spark privacy concerns

Effective enforcement may require identity verification for all users, not just minors, raising anonymity concerns. Proposals reference systems like FranceConnect and future EU digital wallets to confirm age. Experts warn this could centralize sensitive data and expand state visibility into online activity. Technical options like facial age estimation remain unreliable near age thresholds.

Kimi K3 challenges Western leaders

Moonshot AI unveiled Kimi K3 with 2.8 trillion parameters, rivaling top models like Claude Opus 4.8 and GPT-5.6. Benchmarks place it near leading systems, including strong performance on real-world coding tasks. Its rapid jump in rankings underscores accelerating progress from Chinese labs. The release intensifies competition at the frontier of large models.

Open pricing pressures OpenAI, Anthropic

Kimi K3 is priced around $3/M input tokens and $15/M output tokens, undercutting premium offerings. Its open-weight approach lets enterprises run the model independently, reducing reliance on closed APIs. This threatens pricing power for firms like OpenAI and Anthropic. The shift could accelerate a broader move toward self-hosted, agent-oriented deployments.

AI hedge fund hit, Citadel deal

Situational Awareness LP, led by Leopold Aschenbrenner, reported a -67% July drawdown amid AI stock volatility. The firm executed a block sale of public equities to Citadel, eliminating leverage and stabilizing liquidity. Despite the shock, it remains operational and up roughly 80% YTD. The episode highlights crowded trades and reflexive dynamics in AI-linked markets.

ChatGPT tiers split on privacy

ChatGPT Plus (~$20/month) now sits alongside higher tiers like Pro and Business with expanded capabilities. Pro offers up to 20× more tokens and advanced models, with pricing ranging roughly €103–€229/month. Business emphasizes governance, disabling training on user data by default and adding admin controls. Usage accounting differs between Work and Chat modes, affecting cost optimization.

AI “executives” reach 60% viability

Agent-based systems can replicate about 60% of executive functions such as a Chief Marketing Officer. They analyze data, generate strategies, and monitor performance continuously, but still require human oversight. Implementations work best as augmentation rather than full automation. Frameworks like SEED and structured, instruction-driven agents improve reliability and repeatability.

Fable 5 beats Soul 5.6

In a timed build on Higgsfield, Fable 5 produced a playable escape-room game in about 15 minutes. Soul 5.6 took roughly 54 minutes, with slower generation and iteration cycles. The Fable output showed stronger prompt adherence, cohesive design, and faster usability testing. Results suggest latency and execution consistency are now decisive factors in coding model adoption.

Videos covered

Previous briefings · AI